Lesson 3 · AI Foundations Free ~7 min read Decision framework + use cases

Which AI for which job — the decision framework.

There's no universal best AI. There's the right AI for the task in front of you. ChatGPT, Claude, Gemini, Perplexity, Copilot, and Grok are not interchangeable — each is shaped for a different kind of work. This lesson is the decision framework most people skip: a 10-second triage that picks the right tool, plus the use cases each one wins by a wide margin.

The mental model

"Which AI is best?" is the wrong question.

The right question is: What kind of work am I doing right now? Asking "which AI is best" is like asking "which kitchen knife is best." A chef's knife is best for chopping vegetables. A bread knife is best for slicing bread. A paring knife is best for peeling an apple. If you only own one knife, you'll be fine for most tasks — but you'll struggle on others.

The Reframe

Match the AI to the work, not the other way around. Power users keep 2-3 AIs open and switch between them based on the task. It feels like more friction up front; it saves you 10x the friction of fighting the wrong tool.

The 10-second decision tree

Ask yourself one question first:

1. Do I need current, source-cited facts? → Perplexity (or ChatGPT/Gemini with search enabled). Use case: prospect research, fact-checking, current events, sourcing claims.
2. Am I working inside Microsoft 365 or Google Workspace? → Copilot (M365) or Gemini (Workspace). Use case: drafts inside Outlook/Word/Excel/Teams, or Gmail/Docs/Sheets. The integration is the value.
3. Do I need long-form, thoughtful writing or careful reasoning? → Claude. Use case: long documents, nuanced reasoning, writing that needs to feel human, careful code review.
4. Do I need a custom assistant I'll use many times? → ChatGPT (Custom GPTs) or Claude (Projects). Use case: reusable workflow assistants, role-specific tools, structured tasks you do repeatedly.
5. Do I want commentary on current events with attitude? → Grok. Use case: real-time takes on news, X/Twitter sentiment, things deliberately uncensored.
6. None of the above? → ChatGPT. Default for general-purpose work because of breadth and feature depth.

Where each AI wins

ChatGPT
Wins at: Custom GPTs, brainstorming, image generation via DALL-E, voice mode for hands-free use, agents that handle multi-step tasks. The default if you only have one AI.
Claude
Wins at: long-form writing that doesn't sound AI-generated, nuanced reasoning, careful code review, Projects with custom knowledge bases, Artifacts for code/docs, computer use for browser automation.
Copilot
Wins at: any work inside Microsoft 365 — Outlook drafts, Excel formulas, Word docs, Teams summaries, GitHub code completion. Power comes from integration, not the model.
Gemini
Wins at: Google Workspace integration (Gmail, Docs, Sheets, Calendar), NotebookLM for research with sources, long-context tasks (entire books, codebases), Deep Research.
Perplexity
Wins at: any time you need current, sourced information. Sales prospect research, fact-checking, market analysis. Replaces 80% of your Google searches for research tasks.
Grok
Wins at: real-time X/Twitter integration, current-events commentary, queries where you want fewer guardrails. Niche but real value for specific use cases.

The patterns to avoid

Three common mistakes most users make:

1. Using ChatGPT for current facts. Chat AIs generate from memory; their training data has a cutoff. Use Perplexity (or a search-enabled mode) instead.

2. Using Copilot for general writing outside M365. Copilot's value is the M365 integration. For a generic writing task with no inbox/spreadsheet/doc context, you'd be better served by ChatGPT or Claude.

3. Using one AI for everything because switching is annoying. The 10-second cost of opening a different tab is way smaller than the time you waste fighting the wrong tool on the wrong task.

Try it in the playground

Run this prompt across all five models: "What's the current 30-year fixed mortgage rate in the US, and what's it expected to do over the next 6 months?" Watch which models hedge appropriately (because the answer is time-sensitive) vs. which confidently give you outdated numbers. This is the framework in action.

Open Playground →

Final challenge: catalog your work for a week

For one work week, keep a quick note every time you reach for an AI. Track:

  1. What were you trying to do?
  2. Which AI did you use?
  3. Was it the right tool for the job? (per this framework)
  4. If wrong, what would have been better?

By Friday, you'll know which AIs to keep open by default. Most knowledge workers end up with a stable mix of 2-3 daily tools and skip the rest.

Back to Foundations → Try the Playground →

What you can do now

  • Apply a 10-second decision tree before reaching for any AI
  • Match the AI to the type of work in front of you, not the other way around
  • Know which AI wins at each major category of task
  • Avoid the three most common mismatches that waste people's time
Free
Up next in Foundations

Lesson 4 · When NOT to use AI

The cases where reaching for AI costs you more than just doing it yourself. Recognizing them saves you from looking like the person who lets AI write everything — including the bad parts.